Capacity-oriented approaches to developing childhood obesity interventions: a systematic review

Abstract

Capacity-oriented approaches to health interventions seek to empower the target population or community to manage the health issue themselves using resources they can control. Positive deviance, resilience and asset-based approaches are three such methods for developing and implementing health interventions. This study aimed to review the efficacy of interventions explicitly applying these methods in addressing childhood obesity using adiposity as the primary outcome, measured by standardized body mass index. The search strategy was developed and implemented across four electronic databases. Of the 181 records identified and screened, 11 studies were identified as using a capacity-oriented approach overall. Asset-based approaches (n=8 studies) enrolled 47,880 participants, positive deviance (n=2 studies) enrolled 781 participants, and resilience-based interventions (n=1 study) enrolled 35 participants. The asset-based approaches were mixed with three of the eight studies showing a significant reduction in adiposity while the other five did not find a difference. The positive deviance and resilience-based studies showed signs of efficacy in reducing adiposity. There was significant design heterogeneity across studies and varied interpretations and definitions of the approaches used. Further work should attempt to bring some consensus on the use of these approaches to facilitate comparison and advance the science of capacity-oriented interventions for childhood obesity.

Introduction

Socio-economically disadvantaged populations experience significant disparities in childhood obesity across multiple countries and settings.^1–5 To address the epidemic of childhood obesity, scalable and sustainable approaches are needed, and even if resource-intensive approaches are shown to be effective, implementing them for socio-economically disadvantaged populations may prove challenging. Recent interest in capacity-oriented or positive approaches to the problem of childhood obesity has been in part driven by this paradigm.

There are multiple versions of a capacity-oriented approach described in the literature, each with some overlap with the others both conceptually and pragmatically (Figure 1). The three primary approaches described are resilience-based approaches, a positive deviance/positive outlier approach, and assets-based approaches.^6–8 We define the three approaches below and briefly discuss how they intersect.

Figure 1.

Venn diagram depicting the overlap between the theoretical foundations for the three approaches to obesity intervention described: resilience, positive deviance/outliers, and asset-based

Resilience is a construct, with academic roots in psychology, defined as adapting well to an adverse event or continuing to thrive in the face of stress.⁶ This has been more of an individual-level construct traditionally describing an individual’s success despite their own, relatively uniquely defined, adverse life experiences or stressors.

Positive deviance has traditionally and most actively been used to address the problem of child malnutrition.⁹ The core concept of positive deviance recognizes that some individuals succeed despite having the odds stacked against them.¹⁰ One of the early reviews of the subject as related to malnutrition also cites the concept of resilience as being highly similar.⁷ Recently, some authors have also used the term positive outliers to describe the same method.¹¹ Positive deviance is a more population-level approach that defines individuals who succeed despite an adverse environment wherein the majority fail to succeed.

Asset-based approaches to health interventions or asset-based community development is a more public health-oriented approach that seeks to identify community capacity and potential in contrast to an approach that identifies needs and deficits in a community.⁸ Traditionally it has been used in building communities from a development perspective rather than specifically targeting a specific problem such as childhood obesity. Asset-mapping is a core tool used in applying this approach.

We designed the current study to answer the question: do capacity-oriented approaches reduce adiposity compared with usual care (no explicit intervention) among children with overweight or obesity? Our goal was to identify and describe commonalities and differences between interventions explicitly applying the three approaches (resilience, positive deviance and asset-based) to address childhood obesity.

Methods

This systematic review was registered with PROSPERO International prospective register of systematic review with ID CRD42016052556 and adheres to the PRISMA guidelines for reporting systematic reviews.^12,13

Eligibility criteria

Studies were eligible for inclusion in the review if they explicitly applied any one of the following three concepts in designing the intervention: resilience, positive deviance, or asset-based approaches. Additionally, eligible studies included children (age 0–18 years of age) in the study sample and reported an objective measure of weight status (e.g. body mass index percentile) in the population of interest.

Information sources

We conducted a search of MEDLINE/PubMed (1950 to present), SCOPUS (1966 to present), CINAHL (1937 to present), and PsycINFO (1800s to present) in June of 2017.

Search strategy

Literature search strategies were developed by authors BF and CG to search three specific interventions related to preventing childhood obesity: asset-based approaches, positive deviance, and resilience. Title, abstract, key words, and medical subject headings (MeSH) were searched for these three individual concepts (in combination with childhood obesity) as detailed below. Preliminary searches were evaluated and relevant articles were identified. The reference lists of the articles were hand searched to identify any other articles that may have been overlooked. The literature search was limited to human subjects but no language limits were imposed. The initial MEDLINE/PubMed search was adapted for the 3 other electronic databases.

For the concept of childhood obesity, different combinations of the following search terms were used: childhood obesity, childhood adiposity, childhood weight gain, pediatric obesity, pediatric weight gain, weight gain, obesity, and adiposity. The last three terms were filtered with the age limit: Child: birth - 18 years and with infant OR child OR adolescent.

While asset-based community development was found to be associated with a number of MeSH terms including residence characteristics and social change, incorporating these into a search yielded results that were determined to be out of scope for the purpose of this review. In order to focus exclusively on the specific concept of asset-based public health, a key word search strategy was employed utilizing asset-based OR asset mapping. These two key words were then combined with MeSH terms: Community Health Services, Community Networks, Community-Institutional Relations and keywords: community intervention and community programs. Similarly, a key word strategy for positive deviance made use of the combination of positive deviance OR positive outliers. For the concept of resilience, a search strategy combining the keyword resilience OR MeSH term of psychological resilience was utilized. Each of the concept searches (asset-based community development, resilience, and positive deviance) was then combined with the search for childhood obesity. Duplicates were removed utilizing reference management software.

Study selection

Study screening was facilitated using Covidence software¹⁴ with two authors screening each abstract, two authors screening each full-text article and any conflicts resolved via discussion. Systematic reviews related to the topic were screened for inclusion both overall and, if relevant, their references of studies included were then added to the screening process at the abstract stage if not already included from the original search.

Outcomes and prioritization

The primary outcome assessed is any objective measure of adiposity to include body mass index and its variations: waist circumference, body weight, fat free mass etc. The secondary outcomes are intermediate behavioral outcomes such as diet or activity measures or other behavioral change outcomes intermediate to an adiposity outcome such as screen time, sleep or other influencing behaviors.

Risk of bias in individual studies

Risk of bias was assessed independently by two reviewers. Risk of bias was assessed using a modified version of Cochrane’s risk of bias tool.¹⁵ We modified it to include assessments of baseline imbalance and contamination, two items recommended for assessments of non-randomized interventional studies.

Results

Search results

There were 172 studies screened after removal of duplicates and eleven studies included after applying the inclusion and exclusion criteria (Figure 2). Of the 76 studies assessed for full-text screening, the most common reason for exclusion was wrong study design. These were often cross-sectional studies using one of the conceptual approaches to describe a population but lacking any intervention. Many in the wrong intervention category for exclusion were community-based trials that did not explicitly use one of the three conceptual approaches.

Figure 2.

Flow diagram adapted from PRISMA showing results of search strategy and exclusions at stages of review

The identified interventions (n=11) varied by population and context (Table 1). Children from 0–18 years of age were included in the studies. Six of the studies were completed as part of a larger, regional project in the Pacific, and the other interventions were based in the USA (Texas, Arizona, South Carolina and Massachusetts). The studies were judged to have too much heterogeneity in design for a meta-analysis to be meaningful.

Table 1.

Characteristics of included studies and outcome data for anthropometric measures of adiposity by date of publication

Study ID:	Approach	Study design	Subject characteristics	Inclusion criteria, setting	Interventions compared	Intensity	Outcome data (adiposity measure)
Sanigorski 2008	Asset-based	Repeat cross-sectional, quasi-experimental	Children aged 4–12 years (n=833 intervention, n=974 control)	Children in pre-school or elementary school in two areas of Victoria, Australia	Intervention: five objectives (TV, SSBs, PA, active transport, diet) influenced via social marketing, policies and programs Comparison: usual care	Low intensity overall (exposure hours undefined) but long duration (3 years); variable from changes in policies and training for teachers to an after-school physical activity program, and provision of sports equipment	Between intervention and comparison at follow-up: BMIz: −0.11, 95% CI: −0.21, −0.01) WC: −3.14cm, 95% CI: −5.07, −1.22) Overall obesity incidence rate ratio: 0.91, 95%CI: 0.65–1.28
de Silva Sanigorski 2010	Asset-based	Repeat cross-sectional, quasi-experimental	Children aged 2 and 3.5 years, regardless of weight status (n=2,539 intervention, 34,129 control)	Children living in the intervention or control areas who attended health visits across the state of Victoria, Australia	Intervention: community-wide changes to policy, sociocultural and physical environment in early childhood (<5 years) care an education to increase capacity to promote healthy eating and PA Comparison: usual care	Low intensity overall (exposure hours undefined) but long duration (4 years); variable from changes in policies and training for teachers to posters and brochures to families, and a new active play program	Between intervention and comparison at follow-up: 2 year old BMIz: −0.01, 95% CI: −0.04, 0.01 3.5 year old BMIz: −0.04, 95%CI: −0.7, −0.01 Reduced prevalence of overweight/obesity compared to baseline in intervention (2 year-olds 2.5% and 3.5 year-olds 3.4% lower) vs. comparison (0.7% lower for both age groups), p<0.05
Millar 2011	Asset-based	Cluster RCT	Children aged 13–19 years, (n=1276 intervention (5 schools), n=778 control (7 schools)	Children in secondary schools, Victoria, Australia	Intervention: Community capacity framework where action plans were developed by stakeholders, including principal, teachers, parents, and students Comparison: usual care	Low intensity overall but long duration (2 years), highly variable across schools, policy changes, environmental changes (water fountains, removal of soda from vending machine)	Between intervention and control at follow-up: BMIz: −0.07, SE: 0.03, p=0.03 Body fat percent: −0.23, SE: 0.40, p=0.58 Proportion with overweight/obesity: OR: 0.75, SE: 0.14, p=0.12
Fotu 2011	Asset-based	Interventional cohort	Children aged 11–19 years, (n=815 intervention (7 schools), n=897 controls (6 schools)	Children on two islands in the Kingdom of Tonga	Intervention: Social marketing, capacity-building efforts to target PA, diet, reduce SSBs, sedentary activities Comparison: usual care	Low intensity overall but long duration (3 years), variable from training and workforce development to planting local fruit trees and holding sports events	Between intervention and control at follow-up: BMIz: −0.03, SE: 0.03, p=0.26 Body fat percent: −1.46, SE: 0.21, p<0.001. Proportion with overweight/obesity: OR: −0.05, SE: 0.24, p=0.84
Kremer 2011	Asset-based	Interventional cohort, quasi-experimental	Children aged 13–18 years, (n=897 intervention (7 schools), n=2,069 control (11 schools)	Attending a secondary school in selected areas of Viti Levu, Fiji	Intervention: capacity building within schools partnered with faith-based organizations to target seven different objectives around PA and diet Comparison: usual care	Low intensity overall but long duration (2 years), variable from breakfast promotion and policy changes, walking days and involvement of safety officers, to awareness (pamphlets) on PA	Between intervention and comparison at follow-up: BMIz: 0.02, 95%CI: −0.02, 0.07. Body fat percent: −1.17, 95%CI: −1.73, −0.60 Proportion with overweight/obesity: 0.32, 95%CI: −0.03,0.71, p=0.07
Utter 2011	Asset-based	Repeat cross-sectional, quasi-experimental	Children aged 15–18 years, (n=1023 intervention (4 schools), n=589 control (2 schools)	Attending an included school in South Auckland, New Zealand	Intervention: Student health councils designed and implemented activities to increase PA, reduce SSBs, decrease TV Comparison: usual care	Low intensity overall but long duration (2 years), variable from 1–2 times/week breakfast clubs and redesigned cafeteria menu to week long self-empowerment course for 100 students	Intervention baseline mean BMIz: 1.02, 95% CI 0.9, 1.2, and follow-up 1.11, 95% CI 1.0,1.3. Control baseline mean BMIz: 1.00, 95% CI 0.8, 1.2 and follow-up 0.95, 95%CI: 0.8, 1.5, p=0.13
Siwik 2013	Resilience-based	Randomized, lagged intervention	Children aged 8–11 years, n=35 randomized, n=32 with follow-up data	Child BMI above the 85th percentile, Arizona, USA	Group office visit intervention targeting resilience. Control randomized to waitlist.	Group visit: weekly, 90 minutes per session over 12 weeks	Intervention effect on BMI z-score of −0.046, p=0.017 and for BMI of −0.34, p=0.025
Rushton 2015	Asset-based	Prospective cohort with control	Newborns followed from birth through 15 months (n=51 intervention, n=51 matched controls)	Medicaid insurance, otherwise, South Carolina, USA	Intervention: group well-child visits held at an elementary school and a home visiting program Comparison: usual well-child care in the office	Group well-child visits: 90 minute sessions x 6 from birth-15 months; home visiting program twice a month	Intervention group: 4/51 (8%) were overweight by weight/age >90th percentile and 12/51 (24%) in controls, p=0.03. Intervention group: 7/51 (14%) were overweight by weight/height and 11/51 (22%) in controls, p=0.35
Foster 2016	Positive deviance-based	Pilot RCT	Children aged 2–5 years, n=60 randomized, 48 with follow-up data	Child BMI≥95th percentile in Head Start centers, Texas, USA	Intervention: parent mentor using positive deviance-based strategies Comparison: community health worker providing coaching and education	Parent mentors had two, 1-hour contacts/month for six months, community health worker had one, 1-hour contact/month for six months	Between intervention and comparison at follow-up: BMIz: −0.02, 95%CI: −0.26, 0.22 Pre/post change in BMIz over 6 months: For both groups over intervention, change in BMIz: −0.24, 95%CI: −0.34, −0.15
Peters 2010	Asset-based	Interventional cohort with matched control	Children aged 4–8 years, n=601 intervention, n=358 control	Ontario, Canada	Intervention: complex community intervention with goals of promoting children’s well-being and strengthening the community integrated with local services Comparison: two communities matched on sociodemographics	Low intensity overall though long duration (4 years); variable from child directed programs and after-school activities to parent support groups and community events	Intervention group had higher BMI at grade 3, no significant difference at grades 6 or 9; No adjusted measures of BMI or population level measures of adiposity reported
Taveras 2017	Positive-deviance based	RCT	Children aged 2 to 12 years, n=721 randomized, n=664 with follow-up data	Child’s BMI≥ 85th percentile, 6 primary care practices, Massachusetts, USA	Intervention: Enhanced pediatric primary care (EPC), health coach, text messages and mailings around behaviors of screen time, SSBs, diet, PA, sleep, social-emotional wellness Comparison: EPC plus a Neighborhood Resource Guide, text messages around health behaviors	Intervention: EPC plus 6 visits over a year from a health coach (in-person, phone or videoconference), twice weekly text messages Comparison: EPC plus text messaging or emails (less frequent at ~8 times/year)	Between intervention and comparison at follow-up: BMIz: −0.02 (−0.80 to 0.03) Pre/post change in BMIz over 1 year: control BMIz: −0.06, 95%CI: −0.10, −0.02, and intervention BMIz: −0.09, 95%CI: −0.13 to −0.05

RCT = randomized clinical trial; BMI = body mass index; PA = physical activity; SSBs = sugar-sweetened beverages; WC = waist circumference; BW = body weight; PCP = primary care provider

Resilience results

The search identified one published article (Siwik 2013¹⁶) with an intervention designed around the concept of building resilience. Another article (Chandler 2015¹⁷) was initially identified but on further review was found to have included only adolescents 18 years of age or older and was thus excluded from the primary analysis. In Siwik et al 2013, school-age children and their parents were randomized to either a 12-week group office visit intervention or a wait-list control. The intervention provided education around physical activity, nutrition and decreased TV time in the context of teaching resiliency with the goal “to empower children and parents to recognize their choices.”¹⁶ They found a decrease in BMI z-score with an intervention effect of −0.046 (SE=0.02), p=0.017, and also demonstrated a shift towards higher levels of physical activity.

Positive deviance results

There were two studies with published results using positive deviance to inform their intervention (Foster 2016¹⁸ and Taveras 2017¹⁹). These were both randomized clinical trials (RCTs) with two arms though Taveras et al was a larger study. Notably, neither study included a true, non-intervention control (usual care). In Foster et al 2016, parents and their children were randomized to either a community health worker intervention or a parent mentor trained in the positive deviance strategies identified from the community. In Taveras et al 2017, parents and their children were randomized to either receive a primary care-based intervention described as enhanced care with clinical decision support tools, educational materials and a neighborhood resource guide or enhanced care plus a health coach who used motivational interviewing techniques and strategies derived from focus groups of positive deviants. These focus groups identified five themes that were incorporated into the intervention arm: 1) making changes as a family, 2) setting rules and limits around snacking and screen time, 3) being involved in decision-making with their provider, 4) using immediate outcomes to help motivate, and 5) effectively using community resources. In Foster et al 2016, five strategies identified using a positive deviance approach were incorporated into the intervention arm as well. In that study, these were: “dealing with behavior problems without using food, identifying internal motivators for healthy habits, organizational strategies for feeding, accurate perceptions of weight, and effective snacking strategies.”¹⁸

In both of these studies, significant reductions in adiposity were found in both arms of the study when compared with the baseline measures (Table 1). However, neither study found a difference between the intervention (positive deviance) arm and the comparison arm.

Asset-based community development results

Eight studies (Sanigorski 2008²⁰, de Silva-Sanigorski 2010²¹, Fotu 2011²², Kremer 2011²³, Utter 2011²⁴, Millar 2011²⁵, Rushton 2015²⁶, Peters 2016²⁷) were characterized as having used an asset-based approach in the design and implementation of their intervention as published. All of the studies identified and characterized by the research team as fitting the inclusion criteria for this approach used an approach to build off of existing resources in the community and specifically focused on developing capacity from those assets.

Six of the studies (Sanigorski 2008²⁰, de Silva-Sanigorski 2010²¹, Fotu 2011²², Kremer 2011²³, Utter 2011²⁴, Millar 2011²⁵) were all part of one larger project, the Pacific Obesity Prevention in Communities project²⁸ which used a “community capacity-building approach” in the design and implementation of their interventions. As outlined in Table 1, there were mixed outcomes from these studies with three not showing a significant reduction in adiposity^22–24, one study having mixed results²⁵, and two demonstrating positive effects on adiposity (reducing adiposity).^20,21 All of the studies utilized schools as part of the community assets leveraged to conduct the intervention. One of the notable distinctions between the successful and unsuccessful trials from this project was the age of the children. The two studies showing a consistent effect involved much younger children.^20,21 As noted in Table 1, these studies were all relatively low-intensity for the individual in terms of exposure with more policy-level and training of teachers and students types of activities conducted over years.

In Rushton et al 2015²⁶, the authors describe a program that identified existing community assets and then used those assets to provide a health promotion program for low-income parents of infants. They used a combination of group well visits and a home visitation program to promote health and safety under the overarching goal of preventing child abuse. While this intervention was not specifically targeted at reducing obesity, they found a lower proportion of children in the intervention group compared with a matched control to be overweight at 15 months of age, p<0.03 (Table 1).

Peters et al 2010²⁹ similarly used existing assets to build capacity in the community and did not directly target childhood obesity. This community-based program run in Ontario, Canada in the 1990s sought to promote child health with projects that used existing community resources and was community-driven in terms of decision-making. They assessed height and weight after children completed up to four years of participation at grade 3 and then also at grades 6 and 9. They found that the intervention was associated with a higher BMI (18.1 v 17.4 kg/m², p<0.01) compared with matched controls at grade 3 and then no difference seen at further follow-up time points. Notably, proportion overweight or obesity was not reported and the mean BMIs reported are at approximately the 75^th percentile.

Risk of bias assessment

As outlined in Table 2, the non-randomized nature of many of the studies required judgment of a high risk of bias in sequence generation and allocation concealment. Risk of bias due to blinding was judged to be low for most studies given the primary objective measure of adiposity assessed. The risk of bias in incomplete outcome data was judged to be high for many studies with significant drop-out rates, and there were also issues with baseline imbalance across many of the assets-based interventions (Table 2). The self-reported nature of many of the secondary outcomes assessed across studies makes them vulnerable to social desirability bias in the parents completing them, but this should not affect the primary outcome of adiposity. The un-blinded nature of the assessors for many of these secondary outcomes also carries a high risk of bias that did not affect the primary outcome of adiposity.

Table 2.

Risk of bias assessment for the primary outcome of adiposity measure using a version of the Cochrane risk of bias assessment tool, modified for controlled before and after studies to include baseline imbalance and contamination. Other sources of bias and areas unclear clarified in footnotes

Study ID:	Random sequence generation	Allocation concealment	Blinding of participants and personnel	Blinding of outcome	Incomplete outcome data	Selective reporting	Baseline Imbalance	Contamination	Other sources of bias
Sanigorski 2008	High	High	Low	Low	Low	Low	Low	Low	High1
de Silva Sanigorski 2010	High	High	Low	Low	High	Low	High	Unclear4	High2
Millar 2011	High	High	Low	Low	High	Low	Low	Low	Low
Fotu 2011	High	High	Low	Low	High	Low	High	Low	Low
Kremer 2011	High	High	Low	Low	High	Low	High	Low	Low
Utter 2011	High	High	Low	Low	High	Low	Unclear5	Low	High
Siwik 2013	Unclear6	Low	Low	Low	Low	Low	Low	Low	Low
Rushton 2015	High	High	Low	Low	Low	Low	Low	Low	High3
Foster 2016	Low	Low	Low	Low	Low	Low	Low	Low	Low
Peters 2010	High	High	Low	Low	High	Low	Low	Low	Low
Taveras 2017	Low	Low	Low	Low	Low	Low	Low	Low	Low

¹Selection bias possible given non-random selection of community for intervention that wanted to engage in the capacity-based approach

²Selection bias possible given reporting through health centers and may be more likely to select for more educated or higher income population who complete physical examinations

³Retrospective matching of the comparison sample carries high risk of bias as only comparison families with follow-up selected

⁴Given the repeat, cross-sectional design and both intervention and control occurring in the same region of Australia, it was unclear how much risk of contamination would exist

⁵While intervention and comparison groups balanced across most characteristics, not enough information provided to complete judgement

⁶Not enough information provided within the manuscript

Discussion

The conceptual overlap between resilience, a positive deviance/positive outlier approach, and asset-based approaches is that each one recognizes the capacity of a community or individual to succeed. They differ largely on the level at which they operate from a socio-environmental framework. Resilience has its underpinnings in the individual’s capacity to overcome, positive deviance in strategies some individuals use that make them outliers in their community, and asset-based in leveraging the community resources. All of these approaches also require some assessment of the social and environmental context when implemented. In this systematic review, we identified interventional studies for childhood obesity using each of these approaches.

The lone resilience-based intervention demonstrated a statistically significant reduction in adiposity.¹⁶ This approach has strong theoretical underpinnings in empowerment leading to greater self-efficacy. Measuring mediators such as empowerment and self-efficacy along with long-term outcomes are needed to further evaluate this approach.

The positive deviance-based interventions showed no difference from their comparison in adiposity outcomes.^18,19 As the comparison groups also received an intervention of similar intensity in both studies, it is difficult to ascertain these interventions’ true efficacy. In the application of positive deviance to malnutrition, the assessment of long-term outcomes demonstrated the true value of a positive deviance-based approach.⁹ The inherent sustainability of using a solution derived from successful outliers in the community was demonstrated by sustained success in the intervention group. As the positive deviance approach to obesity develops further, using true controls and measuring longer-term outcomes are needed.

Of the assets-based approaches identified, there were mixed results overall with most studies finding no significant intervention effect. As compared with the interventions described under positive deviance and resilience above, these interventions were much less intense in terms of individual contact between participants and intervention staff. The notable exception is the study by Rushton et al, which was similar in terms of intensity to the positive deviance and resilience studies and did see a modest effect in adiposity outcome. Ruston et al is also notable among the assets-based approaches in having the most focused effort on individual behavior coaching and intervention.

One of the challenges of conducting this systematic review was in applying the definition of an assets-based approach to interventional studies around obesity. There were a number of studies that used either community-based participatory research or other community engagement strategies to guide the intervention development and implementation.^30,31 This explains the relatively high proportion of articles that required full-text screening. Ultimately, only articles that explicitly used and described either an asset-mapping approach or capacity-oriented approach in their design were selected for inclusion. It is recognized that there are other community-based interventional studies which may well boost the capacity of the community to address obesity.³¹ However, there is a critical distinction between coincidentally using a theoretical approach to a problem compared with explicitly designing an intervention around a given theory.

Another limitation is publication bias related to significant outcomes. Especially given the nascent nature of inquiry using these approaches to the problem of obesity, it is possible that smaller, pilot studies with null findings exist and have not been published. One limitation may be the specificity of the terms used in the search as it is possible that there are studies that applied a resilience-based approach without including resilience as a keyword or indexed differently. Based on the experience in developing and testing the search terms outlined, we think this is unlikely to be the case for a study that intentionally used one of these approaches.

Using a common language to describe the design of interventions which take an explicit capacity-oriented approach in their design is one tangible step identified by this review that would potentially help move the field forward. This systematic review identifies some of the strengths and limitations of using a capacity-oriented approach via the qualitative synthesis. Unfortunately, one of the theoretical strengths of sustainability is not measured in most of the studies identified given lack of long-term data collection.

Of the three approaches described, resilience and positive deviance/outlier approaches seem to have the most evidence of efficacy in reducing adiposity among children though the evidence is modest and limited by the number of studies published. The lack of evidence supporting an asset-based approach to intervention may be secondary to the relative low-intensity of the interventions published. Capacity-oriented approaches have theoretical potential, particularly for low-resource settings and reducing disparities; however, they will need to be tested in more communities and arguably using true controls to better evaluate their potential.